Los celos y la desconfianza son violencia en el noviazgo

4.1: Summary and main findings

A key role of transcription factor p63 in mediating epidermal commitment, development, and differentiation has been extensively demonstrated through mouse models and causative point mutations in human disease. In mice, knockout of p63 leads to developmental and morphological defects in the squamous epithelia and epidermis, causing abnormal craniofacial development, truncated limbs, and loss of salivary glands, hair follicles, and teeth1,2. In humans, germline missense mutations in p63 cause ectodermal dysplastic syndromes, which are characterized by orofacial clefting and limb malformations3. To date, how p63 engages and modifies chromatin to promote normal development during these early stages is still being elucidated.

Here we report a trans-differentiation model as a novel approach to determine the role of p63 in craniofacial development. We show that p63 can establish epithelial enhancers to regulate downstream genes and that this process is disrupted when p63 carries mutations derived from human disease. In particular, our results show a novel mechanism underlying pathogenic p63 mutations, in which failure to open and establish enhancers leads to transcriptional dysregulation. We integrate our epigenomic data with GWAS datasets derived from nsCL/P patients to identify novel CL/P candidate genes regulated by p63. We also uncover strong enrichment of nsCL/P- associated SNPs at enhancers established by p63, providing mechanistic insight into how these SNPs lead to human pathogenesis. Further, in a follow up study, we identify histone methyltransferase KMT2D as a key partner of p63 at epithelial enhancers. Our data demonstrates for the first time a critical role for KMT2D in epithelial homeostasis by maintaining the adhesion and proliferative capacity of the epidermal basement membrane through its interaction with p63 to

105

regulate critical enhancers for genes involved in both maintaining the basal epidermis as well as in the proper coordination of epithelial stratification.

In our first study4 using an inducible trans-differentiation model converting fibroblasts into keratinocyte-like cells we demonstrated the following main findings:

1. p63 establishes enhancers to upregulate epithelial and inflammation genes: while previous studies had focused on the role of p63’s engagement with chromatin during keratinocyte differentiation5–7, de novo engagement of p63 with chromatin had not been extensively addressed due in part to a lack of tractable human models.

2. Co-binding of p63 and KLF4 leads to robust establishment of enhancers that correspond remarkably well to bone fide enhancers in basal and differentiating keratinocytes: based on published reports8 we expanded our in vitro system to better recapitulate the epigenetic conditions of early development.

3. Binding of p63 to chromatin is a pre-requisite for enhancer establishment: using a patient derived p63 mutation in the DBD we show complete loss of p63 binding to chromatin and loss of enhancer establishment.

4. Binding of p63 to chromatin is required but not sufficient for enhancer establishment: an AEC patient derived substitution mutation in the SAM domain of p63 leads to a protein that retains binding to chromatin but shows defects in enhancer establishment. The main defect is in opening chromatin, pointing towards a potentially novel function of the p63 SAM domain in recruiting chromatin remodelers for enhancer establishment. Further, our results contrast a recent report9 suggesting the main disease mechanism in AEC is aggregation of mutant p63, and posits a novel molecular link to explain disease phenotypes found in these patients.

106

5. We identified potential candidate causative genes for cleft lip/palate regulated by p63 at and outside of known risk loci: by integrating epigenomic data-sets from our trans- differentiation model with GWAS datasets we were able to gain novel insights into molecular mechanisms underlying nsCL/P. In particular, we were able to prioritize candidate genes within half of the known risk loci and also identify over 100 new candidate genes and risk loci for nsCL/P. Importantly, our analysis also has the potential to identify the enhancers that regulate them.

6. Enhancers established by p63+KLF4 are enriched for SNPs associated with nsCL/P: using 2 different co-localization analyses we identify enrichment of SNPs associated with nsCL/P only at newly established enhancers. This thus provides a novel molecular mechanism for how these SNPs may contribute to disease at a molecular level.

Taken together, our findings in our first study address how p63 binds and remodels chromatin with its first encounter, identifies and prioritizes nsCL/P candidate genes and the enhancers that regulate them, and provides new explanations for how p63 mutations and SNPs associated with nsCL/P underlie disease.

In our second study7 we wanted to ask which proteins interact with p63 on chromatin to establish enhancer structure. We identified methyltransferase KMT2D interacting with p63 at a broad array of enhancers involved in epithelial development, adhesion, and differentiation. The main findings in this study were:

1. KMT2D plays a key role in maintaining the hallmark gene expression program and proliferative capacity of epithelial progenitors: KMT2D, one of the most frequently mutated genes in human cancer, has been extensively characterized. However, its role in epithelial cells and homeostais had not been previously explored. We show

107

depletion of KMT2D leads to reduced proliferation and reduced expression of epithelial development and adhesion genes, including DLG2, RARG, ITGB4 and LAMB2.

2. KMT2D is enriched genomewide at p63 target enhancers: ChIP-seq analysis demonstrated that KMT2D and p63 are enriched at overlapping sites across the genome, with as high as 49% of p63 peaks in keratinocytes containing a directly overlapping KMT2D peak. Many of these included known p63 target enhancers, such as MINK1, SERINC2, IRF6 and TP63 itself.

3. KMT2D and p63 interact on chromatin: formaldehyde crosslinking followed by immuno-precipitation and mass spectrometry show p63 and KMT2D interact on chromatin. We also find over 1000 shared binding partners which could mediate this interaction.

4. KMT2D is enriched at p63 targets and maintains the expression of key genes involved in epithelial homeostasis: Knock down(KD) of KMT2D shows reduction of KMT2D at p63 target enhancers and reduction of gene expression of corresponding genes. Pathways affected include epidermal commitment and stratification, as well as cell junction and adhesion.

5. KMT2D can drive expression of p63-target genes through both catalytic histone methylation activity, as well as catalytically independent mechanisms at enhancers: Profiling of H3K4me1 (catalyzed by KMT2D) and H3K27ac (KMT2D recruits the acetyltransferase CBP/p30010–12) shows KD of KMT2D correlates with loss of H3K4me1 and H3K27ac at p63 regulated enhancers of 82 genes involved in development and differentiation. In another 20 loci we observe loss of gene

108

expression with no changes in H3K4me1 and H3K27ac, consistent with recent evidence demonstrating histone methyltransferase independent functions of KMT2D in regulating gene expression13.

6. KMT2D coordinates with p63 to maintain epidermal progenitor gene expression and proper epidermal differentiation and stratification: Investigation of genes upregulated upon KD of KMT2D showed a striking enrichment of genes involved in keratinocyte differentiation, including canonical genes, such as KLF4, GRHL1, GRHL3 and ZNF750. Using a three-dimensional human organotypic skin culture we observed keratinocytes treated with shKMT2D showed disrupted stratification dynamics, as well as premature cornification. Together with the above findings these results demonstrated a key role of KMT2D in maintaining epidermal progenitor gene expression and preventing terminal differentiation.

In conclusion, we provide the first description of KMT2D’s role at epithelial enhancers, including its genome-wide interactions with the transcription factor p63 to coordinate key gene expression programs required for both epithelial cell identity and the maintenance of epithelial progenitors.

For an extended discussion on both studies, please refer to the discussion section of chapter 2 (pages 45-49) and chapter 3 (pages 80-81).

4.2: Future directions

The studies described in chapter 2 and 3 greatly increased our knowledge about p63’s role in epithelial enhancer establishment and revealed a potential disease mechanism in orofacial clefting, connecting SNPs and enhancer regulation. Further, they provided insight into possible binding partners of p63 at enhancers, including histone methyltranferase KMT2D; and described

109

a previously unknown role of KMT2D in maintaining epithelial homeostasis. However, many questions remain. In the following sections I will describe open questions and possible follow up experiments.

4.2.1: Dissecting how specific SNPs may cause CL/P

In chapter 2 we found SNPs highly associated to CL/P are enriched at newly established enhancers by p63 and KLF4. We further discussed in detail two loci that are prominently linked to CL/P but where a molecular mechanism to explain disease has yet to be found: 8q24 locus and MAFB locus14–17. Our colocalization analyses identified in both cases newly established enhancers by p63 and KLF4 colocalize with SNPS highly associated to CL/P, however how these SNPs cause disease remains to be shown. In particular, in these 2 cases and in many other examples we find several enhancers (exhibiting open chromatin and de novo H3K27ac) established by p63 and KLF4 and it remains unclear if all are necessary for downstream regulation of gene expression. Further, in the case of the 8q24 locus and other known risk loci, the exact genes affected by the SNPs have not been determined. To start to answer these questions it would be interesting to test some of these enhancers using luciferase reporter systems to see if they are active in human keratinocytes and our trans-differentiation model, as well as during epidermal development and development of the lip and palate in murine models. SNPs could easily be integrated into these enhancer reporter assays to test their effect in gene expression. A more thorough, albeit technically challenging alternative is to use CRISPR systems to introduce specific SNPs either in our trans differentiation model or an epidermal differentiation model derived from pluripotent stem cells, to test how they affect downstream gene expression and differentiation.

110

A more high-throughput possibility to identify how newly established enhancers regulate downstream gene expression is using chromatin capture assays to study whether these newly established enhancers are looping to promoters. Looping to promoters could then be correlated with differential gene expression to filter key enhancers and corresponding SNPs to test.

Another observation from our studies is that many nsCL/P highly associated SNPs cluster around established enhancers but do not seem to be found within regulatory elements (based on chromatin signature). Whether these SNPs are simply in linkage disequilibrium with disease causative SNPs or actively contribute to disease through independent mechanisms remains to be elucidated. The experiments proposed above would start to shed light into these processes and greatly expand our knowledge around cleft lip/palate pathogenesis.

4.2.2: Understanding the role of lncRNAs in epidermal specification and differentiation Another exciting result from our trans-differentiation experiments is that many lncRNAs become de novo expressed after establishment of enhancers by p63 and KLF4. In particular, in the 8q24 risk locus, which shows the strongest effect size across populations of Asian and European origin18 we observe de novo expression of lncRNA LINC00977. How and whether this and other lncRNAs regulate epidermal specification and differentiation, as well as more complex processes such as lip and palate development is not understood. A recent study harnessed the versatility of gene editing tools to probe how lncRNAs could confer resistance to melanoma treatments19. By using an enzymatically inactive Cas9 (dCas9) version coupled to an activating VP64 protein complex targeted to over 10000 lncRNA promoter regions they were able to screen which ones conferred a resistance phenotype, suggesting a role for specific lncRNAs in this process. To begin to understand the role of lncRNAs in epidermal specification and differentiation, I propose a similar screen. We could target a large amount of lncRNAs in human keratinocytes with either activation (dCas9-VP64) or repression (dCas9-KRAB) while in parallel

111

inducing keratinocyte differentiation using calcium supplemented media. After three days of differentiation, keratinocytes can be sorted using basal vs differentiation surface markers. Comparison of sgRNA counts can then point us towards lncRNAs crucial for keratinocyte differentiation. We could then confirm their function using three-dimensional skin organotyoic cultures as described in chapter 3 and then test their role in development using murine models.

4.2.3: Identifying new SNPs associated to CL/P and CPO

A remarkable finding from the GWAS data on CL/P is that the 40 identified risk loci explain about 25% of the cases of non-syndromic cleft lip/palate20,21. This is particularly surprising given the small sample sizes profiled (between 100-1000 patients per cohort). Most of the data has been collected from patients of Asian and European descent. In order to better understand genetic effects and how non-coding regions can cause orofacial clefting, it is imperative to increase the number of patients sequenced, as well as expanding to diverse non-European populations. Further, while GWAS have provided important insight into CL/P, much more remains unknow about cleft palate only (CPO) pathogenesis. To date, only a single risk loci has been confidently identified to explain CPO22. A major challenge here is the lower incidence of cases, as well as lack of large samples of data collected in these populations.

4.2.4: Understanding discordant phenotypes in identical twins with p63 mutations

A key observation in patients with p63 substitution mutations is that all patients found so far are heterozygous for the mutation and that even patients with the same mutation show very different phenotypes. Given the severe phenotype observed in mice with homozygous deletion of Trp63 (encoding for p63 in mice)1,2 it is likely that complete loss of function of p63 in humans is

112

embryonically lethal. A recent study23 reported the first case of monozygotic twins carrying a p63 mutation in the DBD. Interestingly, the twins exhibited discordant phenotypes, with one twin showing limb malformations and normal craniofacial development and the other showing normal limb development but CL/P23. Together with our studies and the fact that p63 acts as a tetramer, this suggests there is an important dose and spatiotemporal component to p63’s epithelial enhancer establishment during development. Previous studies looking at heterogenous composition of tetramers in p53 found that tetramers composed of at least three mutant p53 proteins showed impairment of transcriptional activity24. It is possible that a similar effect is seen in patients carrying a substitution mutation in one of their p63 alleles. To test this hypothesis, it would be interesting to overexpress wildtype p63 in a cell line that does not endogenously express it. In parallel, mutant versions of p63 could be introduced under a doxycycline inducible promoter and effects on gene expression could be monitored. It would also be interesting to obtain keratinocytes from the monozygotic twins mentioned above and profile active enhancers to see if there are specific differences in enhancer structures. Given the stochastic nature of tetramer formation, it may be possible that defects on enhancer establishment and gene regulation may be lost within a population of cells. A better system would be to utilize a murine model in which a substitution mutation is introduced in a single allele of p63 followed by monitoring diverse phenotypes, more effectively recapitulating the spatiotemporal and dose dependent nature of p63’s enhancer establishment in development.

4.2.5: Translation to treatment to prevent orofacial clefting

An important question is how a better understanding of p63’s role in orofacial clefting can be translated into therapy, especially since CL/P is one of the most common congenital malformations reported, second only to heart congenital malformations20. Firstly, in the case of patients carrying a substitution mutation in p63, patients could possibly benefit from silencing of the mutant allele during development. This idea is bolstered by the observation that mice with

113

homozygous deletion of p63 show severe phenotypes that are lethal shortly after birth, while heterozygous mice with only 1 allele of p63 deleted show no difference compared to mice with two copies of WT p6325. How this would be achieved in human embryo development is presently unclear, a therapy involving allele specific siRNA or a small molecule inhibitor that only recognized mutant versions of p63 are a possibility but must be thoroughly tested for safety and efficiency.

Downstream pathways affected by p63 enhancer establishment may provide less disruptive treatment possibilities to prevent orofacial clefting. For example, we observe p63 establishes enhancers upstream of folic acid receptors FOLR1 and FOLR3 and that establishment of enhancers is followed by de novo expression of the respective genes in our trans-differentiation model. In mice it has been observed that KO of folbp1 (encoding for the homolog of FOLR1 in mice) leads to orofacial clefting26. This phenotype can be mitigated by supplementation of folic acid during embryo development in these mice. Similarly, it may be possible that defects in enhancer establishment in humans due to SNPs at enhancers of folic acid receptors lead to poor upregulation of these receptors and that risk of orofacial clefting might be mitigated through supplementation of folic acid in families that carry these SNPs.

4.2.6: Molecular mechanism behind birth defects observed with thalidomide treatment Thalidomide, a drug used in the 1950s to treat morning sickness in pregnant women lead to over 10.000 infants with birth defects in Europe27. For almost 60 years, the molecular mechanisms behind the resulting birth malformations were poorly understood. A recent study showed degradation of SALL4 as an important pathway in thalidomide induced birth defects27. Remarkably, the phenotypes observed in newborns, included limb and craniofacial malformations (e.g. CL/P) and closely mirrored the developmental phenotypes observed in patients carrying mutations in p63. Because of the high similarities in the phenotypes it is possible that thalidomide

114

interferes in epithelial lineage specification and enhancer establishment via p63. To test whether this is the case, it would be interesting to supplement cells in the fibroblast to keratinocyte-like trans-differentiation model with thalidomide at different doses and see if enhancer establishment is dysregulated. Presently, thalidomide continues to be used to treat a diverse array of diseases, from erythema nodosum leprosum to multiple myeloma. Treatment is correlated with significant adverse effects in the central nervous system, cardiovascular system and the dermis28. Understanding whether thalidomide interferes with enhancer establishment could provide important understanding of adverse effects associated with this drug.

4.2.7: p63’s role in cancer and other disorders

While p63 initially garnered huge interest due to the high degree of sequence identity compared to p53, it was not shown to play a similar role in tumor suppression25. However, it has been observed that p63 is downregulated in some cancers undergoing epithelial to mesenchymal